Here the veto sits with the graphics card, not the power supply. ThermalProtect does not open a contactor and does not trip the supply's own protection. The sensor is wired into the 12V-2x6 sense pins, and at temperature it pulls those lines so the card shuts itself down [4][5]. The cable makes a request and the graphics card carries it out. Corsair puts that at milliseconds, and what the owner sees is a black screen [5]. A black screen looks identical whether the cause is a driver crash or a genuine protection trip, so the likely first response is a reboot and another attempt.
Then the geometry. The ThermalProtect lead is 650mm and the measuring point sits 30mm back from the graphics card connector [15][4], which leaves roughly 95 percent of the run on the far side of the sensor [2]. One sensor, one location, at the end where the heat is expected to appear. For a 65C trip to fire before a terminal is in trouble, heat from a degraded contact has to conduct back into the bundle quickly enough to lift the comb, and the offset between contact face and comb has to hold across seating conditions and cable dress. That is a defensible engineering assumption, and I would take the sensor over a warning leaflet. It is also precisely the figure a review with a thermocouple on the connector face could publish. Until someone does, 65C is a threshold set against Corsair's test connector, not yours.
The commercial shape matters as much as the physics. The monitored lead is rated 600W on a unit rated 1000W, so about 60 percent of the supply's output can pass through the sensed path [1]. Everything else leaves the supply unwatched. Meanwhile the box ships an AC cord, four mounting screws, a handful of zip ties and a booklet, with no Velcro straps, no spare combs and no tester, which Tom's Hardware calls stingy at the price [11]. That fits the stingy pattern already noted: the headline feature lives inside a comb, yet the box holds none in reserve.
Whether the 65C number transfers to your build depends on ambient temperature and on how the graphics card responds to pulled sense lines. The unit is specified for 50C ambient [6], and a case running near that will not behave like a 22C bench. And the shutdown depends on the card acting on pulled sense lines, so a card that responds slowly, or not at all, gets nothing from the sensor [5]. On the parts a reviewer can measure with instruments, the verdict is strong: efficiency, ripple and regulation all good, fan off up to 500W [13][7]. The millisecond trip time is a manufacturer figure in the material at hand [5], and the platform underneath is HEC's RMe, revised [1].
